Forum: Ruby It's a Bird, it's a plane, no! umm, it's Super Boid?

Hey Guys,

I have something actually working that seems allot like Swarm
Technology to me.   The system features many smaller parts that seem
allot like "Boids" (or "Agents").

I'd be tempted to call them "Dorts" because they get thrown straight
at the target and fire on my machine at about 2,000 per second (in
1.9).  Not very "natural" - one would not want to have mosquitoes able
to do that :)

There's no "in-flight" communication between the Boids - their life
cycle involves trying to attain values for a list of required
attributes and a docking with a "HelperClass" (see Ruleby also) for
any in-depth calculations.  They make their assertions, record all the
results, and wait to be debriefed after they have landed.

So are these things Boids or not?

For example:

They are fired sequentially so they are not buzzing around sending
signals to each other.

However, if a Boid is assigned a lower pass number it can "leave
information" for a Boids fired on later passes - so there is
inter_boidal_communications :) - but only after their "flight" /
period of activity.


It's taken me almost 3 years to get this all this far along  Right
now, however, I can go on localhost  and author 10 new real almost
guaranteed to work "Boids" easily in an hour.(When I started and did
these by hand they could take a full day.)   Mostly all I have to do
now
is make  multiple choices.  (Lots of Ruby reflection at work!)

I'm guessing that most you folks here are actively developing Web
applications these days?

Does anybody here happen to follow "Swarm Intelligence"?

Anyway, the power of this technique just seems to keep revealing
itself more and more.   For example, the system now supports "multi-
pass" so that pass 1 Boids can do things based on pass 0 Boids which
is defined as "System".  For another example, the Boids can be
promoted or demoted based on some statistical work we do (related to
their category distribution).

The whole power of all this hits me between my eyes when I make
changes to the UrBoid - the grand-daddy of all the Boids - a few line
changes here and 100s, someday maybe 10,000s of these things can be
taught new behavior.  Fun stuff!

Cheers,

Thunk

This is so far over my head that I thought it was parody for a while. 
Can
you give a link to a demonstration or some resource for the more 
ignorant of
us? It sounds interesting, I just have no base of knowledge to 
comprehend it
from.

thunk wrote:
> Hey Guys,
> 
> I have something actually working that seems allot like Swarm
> Technology to me.   The system features many smaller parts that seem
> allot like "Boids" (or "Agents").
> 
> I'd be tempted to call them "Dorts" because they get thrown straight
> at the target and fire on my machine at about 2,000 per second (in
> 1.9).  Not very "natural" - one would not want to have mosquitoes able
> to do that :)
> 
> There's no "in-flight" communication between the Boids - their life
> cycle involves trying to attain values for a list of required
> attributes and a docking with a "HelperClass" (see Ruleby also) for
> any in-depth calculations.  They make their assertions, record all the
> results, and wait to be debriefed after they have landed.
> 
> So are these things Boids or not?
> 
> For example:
> 
> They are fired sequentially so they are not buzzing around sending
> signals to each other.
> 
> However, if a Boid is assigned a lower pass number it can "leave
> information" for a Boids fired on later passes - so there is
> inter_boidal_communications :) - but only after their "flight" /
> period of activity.
> 
> 
> It's taken me almost 3 years to get this all this far along  Right
> now, however, I can go on localhost  and author 10 new real almost
> guaranteed to work "Boids" easily in an hour.(When I started and did
> these by hand they could take a full day.)   Mostly all I have to do
> now
> is make  multiple choices.  (Lots of Ruby reflection at work!)
> 
> I'm guessing that most you folks here are actively developing Web
> applications these days?

Well, I am, but lots of people here are not.  You'd have better luck in 
the Rails group.  Anyway, what does this question have to do with the 
rest of your post?

> 
> Does anybody here happen to follow "Swarm Intelligence"?
> 
> Anyway, the power of this technique just seems to keep revealing
> itself more and more.   For example, the system now supports "multi-
> pass" so that pass 1 Boids can do things based on pass 0 Boids which
> is defined as "System".  For another example, the Boids can be
> promoted or demoted based on some statistical work we do (related to
> their category distribution).
> 
> The whole power of all this hits me between my eyes when I make
> changes to the UrBoid - the grand-daddy of all the Boids - a few line
> changes here and 100s, someday maybe 10,000s of these things can be
> taught new behavior.  Fun stuff!
> 
> Cheers,
> 
> Thunk

And your point is...?  And the question you were asking is...?

Best,
-- 
Marnen Laibow-Koser
http://www.marnen.org
marnen@marnen.org

thunk wrote:
>
> They are fired sequentially so they are not buzzing around sending
> guaranteed to work "Boids" easily in an hour.(When I started and did
> itself more and more.   For example, the system now supports "multi-
> Cheers,
>
> Thunk
>   

Interesting. Do you have an application in mind for this or is it just
for fun?

For other readers, the Wikipedia entry for "boid" may provide an
explanation of of this general subject.

-- Bill

On Feb 9, 2010, at 05:26 , Marnen Laibow-Koser wrote:

> And your point is...?  And the question you were asking is...?

Your entire response was uncalled for. Thunk's mail was on-topic and 
interesting and I hope to see more from him/her. I can't say the same 
about yours.

Ryan Davis wrote:
> On Feb 9, 2010, at 05:26 , Marnen Laibow-Koser wrote:
> 
>> And your point is...?  And the question you were asking is...?
> 
> Your entire response was uncalled for. 

I don't know what the point of Thunk's post was.  I don't know what it 
had to do with Ruby.  I was asking to try to get some further 
information.  That's never uncalled-for.

> Thunk's mail was on-topic and 
> interesting and I hope to see more from him/her. I can't say the same 
> about yours.

And yet you posted a message with what appear to be thinly veiled 
insults to me for asking a reasonable question.  That doesn't really 
belong here.

Best,
-- 
Marnen Laibow-Koser
http://www.marnen.org
marnen@marnen.org

On Tue, Feb 9, 2010 at 8:21 PM, Marnen Laibow-Koser <marnen@marnen.org> 
wrote:
> Ryan Davis wrote:
>> On Feb 9, 2010, at 05:26 , Marnen Laibow-Koser wrote:
>>
>>> And your point is...?  And the question you were asking is...?
>>
>> Your entire response was uncalled for.
>
> I don't know what the point of Thunk's post was.  I don't know what it
> had to do with Ruby.  I was asking to try to get some further
> information.  That's never uncalled-for.

I didn't quite see the relevance, either, but it was interesting and I
was happy to read it.  I do wish it came with some examples or a
github link.

-greg

... please let us play with this ;)

Somebody had to say it

-- kaspar

Do you have an application in mind for this or is it just
for fun?


The idea is definitely "commercial".   My original thinking was to
keep "the  engine" - where the model is built proprietary, and make
pretty much everything else open source.  The engines would be running
and make building new domains almost trivial - people could
concentrate on build sets of Boids rather than Selection/Navigation
database issues and all of that.  We are using couchDB and it opened
the door to saving the Boids as "documents".

The project was conceived for "checking computer components" as they
are added into a "builder box" or shopping basket if you will.   The
assumption is that all the components are part of a computer system.
This came from my hesitation to hit the Buy button with over $2k worth
of items sitting in my basket at compUSA (to build a Hackintosh). The
web is great at serving up blobs of pretty much self-consistent text
but trying to make sense of something like building a Hackintosh is
not so simple because you don't know who to trust and things are
changing pretty fast. (BTW I installed linux on it and never tried OSX
- but the machine has been rock solid :) fast, quiet,  and about 1/2
the price..  they are just PCs)

(Let me explain: I'm a 30 year programmer dude - haven't "worked"
since '76 when I put a H8 together my first app for McGraw-Hill sold
over $7M the first season back in 1984.  I left programming for awhile
because I was genuinely alergic to c++.  Went off to a beautiful but
lonely resort area here in NE Wis (Door County) and studied
Smalltalk.)

Anyway,  the "Boids" are really simple and generic.  Each one
corresponds to something like a "item" in a check list.  The
attributes and everything it needs are designed to be handed to these
things quickly and in already "standardized" form for units and
such.

The Boid goes on to make classical assertions and builds a detailed
record of its experiences.  It usually uses one Helper Library (a
current restriction).  If multiple "helperLibraries are needed it can
use the data from a Boids in a previous pass(es). Like I say I have
been trying to keep the Boids as simple as possible so they can be
authored by non-programmers.

Because the Boids are so simple to write we see potential for many
domains where there is a lot of confusion due to rapid change, or
where there is a large body of technical knowledge (as medical,
medicines).

The powerful thing using the web would be that, for the medical stuff,
Doctors all over the world could submit "candidate" boids that could
get reviewed and promoted to a national system, and like that.

On a much more humble scope there could be offered a service to check
out your product basket at CompUSA for $1 or so - seems like it would
be worth it as a single missing or wrong item can delay things a week
easy, or can even turn that new gamer dream machine into a big
expensive nightmare.

So, we see all kinds of potential.   I am working on a on-line demo
system and that is just going slow because I'm struggling daily with
web stuff like I cannot believe.  I'm getting some great help when I
need it, however.  Some sort of on-line demo should be available
before too long.

(I am starting to write to some universities with programs for AI
technology transfer hoping to find some interest.)

Cheers,

Thunk

On Feb 10, 2010, at 09:30 , thunk wrote:

> The idea is definitely "commercial".   My original thinking was to keep "the  engine" - where the model is built proprietary, and make pretty much everything else open source.

I have to admit, I'm a lot less interested now. :(

> ...
> The project was conceived for "checking computer components" as they are added into a "builder box" or shopping basket if you will.

And even less so. I don't see how this ties into the idea of having AI 
flocking algorithms figure stuff out for you.

> The assumption is that all the components are part of a computer system. This came from my hesitation to hit the Buy button with over $2k worth of items sitting in my basket at compUSA (to build a Hackintosh). [...] (BTW I installed linux on it and never tried OSX - but the machine has been rock solid :) fast, quiet, and about 1/2 the price..  they are just PCs)

Um. Macs don't cost $4k unless you're building a monster. A quick look 
on apple shows that for $4k I can get 8 cores, 8 gigs of ram and 2.5 Tb 
of disk. OR... I can get a mac mini for $600 or a laptop for 1k. Much 
less than $2k in both cases. Again, no need for AI here to configure or 
purchase. :P

Anyhow. I'd love to see what you're doing, but not if it is going to be 
some proprietary system that I can't play with.  Looks like you took all 
the swarm stuff down. Pity. Based on the gists you have up on github, 
I'd say you have a ways to go with ruby before it'll start feeling right 
(and therefore be accepted by the community)

BTW, how do boids behavior compare to David Gelernter's parallel recipes 
in Linda?